Hongmin Deng

The graph convolutional network (GCN) has become a mainstream technology in skeleton-based action recognition since it was first applied to this field. However, previous studies often overlooked the pivotal role of heuristic model initialization in the extraction of spatial features, impeding the model from achieving its optimal performance. To address this issue, a lightweight initialization-enhanced adaptive graph convolutional network (LI-AGCN) is proposed, which effectively captures spatiotemporal features while maintaining low computational complexity. LI-AGCN employs three coordinate-based input branches (CIB) to dynamically adjust graph structures, which facilitates the extraction of informative spatial features. In addition, the model incorporates a lightweight and multi-scale temporal module to extract temporal feature, and employs an attention module that considers the temporal, spatial, and channel dimensions simultaneously to enhance key features. Finally, the performance of our proposed model is evaluated on three large-scale public datasets: NTU RGB+D, NTU RGB+D 120, and UAV-Human. The experimental results demonstrate that the LI-AGCN achieves excellent comprehensive performances on these datasets, especially obtaining 90.03% accuracy on the cross-subject benchmark of the NTU RGB+D dataset with only 0.18 million parameters, showcasing the effectiveness of the model.

Automated segmentation algorithms for dermoscopic images serve as effective tools that assist dermatologists in clinical diagnosis. While existing deep learning-based skin lesion segmentation algorithms have achieved certain success, challenges remain in accurately delineating the boundaries of lesion regions in dermoscopic images with irregular shapes, blurry edges, and occlusions by artifacts. To address these issues, a multi-attention codec network with selective and dynamic fusion (MASDF-Net) is proposed for skin lesion segmentation in this study. In this network, we use the pyramid vision transformer as the encoder to model the long-range dependencies between features, and we innovatively designed three modules to further enhance the performance of the network. Specifically, the multi-attention fusion (MAF) module allows for attention to be focused on high-level features from various perspectives, thereby capturing more global contextual information. The selective information gathering (SIG) module improves the existing skip-connection structure by eliminating the redundant information in low-level features. The multi-scale cascade fusion (MSCF) module dynamically fuses features from different levels of the decoder part, further refining the segmentation boundaries. We conducted comprehensive experiments on the ISIC 2016, ISIC 2017, ISIC 2018, and PH2 datasets. The experimental results demonstrate the superiority of our approach over existing state-of-the-art methods.

Graph convolutional networks (GCNs), which extend convolutional neural networks (CNNs) to non-Euclidean structures, have been utilized to promote skeleton-based human action recognition research and have made substantial progress in doing so. However, there are still some challenges in the construction of recognition models based on GCNs. In this paper, we propose an enhanced adjacency matrix-based graph convolutional network with a combinatorial attention mechanism (CA-EAMGCN) for skeleton-based action recognition. Firstly, an enhanced adjacency matrix is constructed to expand the model’s perceptive field of global node features. Secondly, a feature selection fusion module (FSFM) is designed to provide an optimal fusion ratio for multiple input features of the model. Finally, a combinatorial attention mechanism is devised. Specifically, our spatial-temporal (ST) attention module and limb attention module (LAM) are integrated into a multi-input branch and a mainstream network of the proposed model, respectively. Extensive experiments on three large-scale datasets, namely the NTU RGB+D 60, NTU RGB+D 120 and UAV-Human datasets, show that the proposed model takes into account both requirements of light weight and recognition accuracy. This demonstrates the effectiveness of our method.

For skeleton-based action recognition, graph convolutional networks (GCN) have absolute advantages. Existing state-of-the-art (SOTA) methods tended to focus on extracting and identifying features from all bones and joints. However, they ignored many new input features which could be discovered. Moreover, many GCN-based action recognition models did not pay sufficient attention to the extraction of temporal features. In addition, most models had swollen structures due to too many parameters. In order to solve the problems mentioned above, a temporal feature cross-extraction graph convolutional network (TFC-GCN) is proposed, which has a small number of parameters. Firstly, we propose the feature extraction strategy of the relative displacements of joints, which is fitted for the relative displacement between its previous and subsequent frames. Then, TFC-GCN uses a temporal feature cross-extraction block with gated information filtering to excavate high-level representations for human actions. Finally, we propose a stitching spatial–temporal attention (SST-Att) block for different joints to be given different weights so as to obtain favorable results for classification. FLOPs and the number of parameters of TFC-GCN reach 1.90 G and 0.18 M, respectively. The superiority has been verified on three large-scale public datasets, namely NTU RGB + D60, NTU RGB + D120 and UAV-Human.

In recent years, great achievements have been made in graph convolutional network (GCN) for non-Euclidean spatial data feature extraction, especially the skeleton-based feature extraction. However, the fixed graph structure determined by the fixed adjacency matrix usually causes the problems such as the weak spatial modeling ability, the unsatisfactory generalization performance, the excessively large number of model parameters, and so on. In this paper, a spatially adaptive residual graph convolutional network (SARGCN) is proposed for action recognition based on skeleton feature extraction. Firstly, the uniform and fixed topology is not required in our graph. Secondly, a learnable parameter matrix is added to the GCN operation, which can enhance the model’s capabilities of feature extraction and generalization, while reducing the number of parameters. Therefore, compared with the several existing models mentioned in this paper, the least number of parameters are used in our model while ensuring the comparable recognition accuracy. Finally, inspired by the ResNet architecture, a residual connection is introduced in GCN to obtain higher accuracy at lower computational costs and learning difficulties. Extensive experimental on two large-scale datasets results validate the effectiveness of our proposed approach, namely NTU RGB+D 60 and NTU RGB+D 120.

As an important basis of clinical diagnosis, the morphology of retinal vessels is very useful for the early diagnosis of some eye diseases. In recent years, with the rapid development of deep learning technology, automatic segmentation methods based on it have made considerable progresses in the field of retinal blood vessel segmentation. However, due to the complexity of vessel structure and the poor quality of some images, retinal vessel segmentation, especially the segmentation of Capillaries, is still a challenging task. In this work, we propose a new retinal blood vessel segmentation method, called multi-feature segmentation, based on collaborative patches. First, we design a new collaborative patch training method which effectively compensates for the pixel information loss in the patch extraction through information transmission between collaborative patches. Additionally, the collaborative patch training strategy can simultaneously have the characteristics of low occupancy, easy structure and high accuracy. Then, we design a multi-feature network to gather a variety of information features. The hierarchical network structure, together with the integration of the adaptive coordinate attention module and the gated self-attention module, enables these rich information features to be used for segmentation. Finally, we evaluate the proposed method on two public datasets, namely DRIVE and STARE, and compare the results of our method with those of other nine advanced methods. The results show that our method outperforms other existing methods.

In recent years, great progress has been made in the recognition of skeletal behaviors based on graph convolutional networks (GCNs). In most existing methods, however, the fixed adjacency matrix and fixed graph structure are used for skeleton data feature extraction in the spatial dimension, which usually leads to weak spatial modeling ability, unsatisfactory generalization performance, and an excessive number of model parameters. Most of these methods follow the ST-GCN approach in the temporal dimension, which inevitably leads to a number of non-key frames, increasing the cost of feature extraction and causing the model to be slower in terms of feature extraction and the required computational burden. In this paper, a gated temporally and spatially adaptive graph convolutional network is proposed. On the one hand, a learnable parameter matrix which can adaptively learn the key information of the skeleton data in spatial dimension is added to the graph convolution layer, improving the feature extraction and generalizability of the model and reducing the number of parameters. On the other hand, a gated unit is added to the temporal feature extraction module to alleviate interference from non-critical frames and reduce computational complexity. A channel attention mechanism based on an SE module and a frame attention mechanism are used to enhance the model’s feature extraction ability. To prevent model degradation and ensure more stable training, residual links are added to each feature extraction module. The proposed approach was ultimately able to achieve 0.63% higher accuracy on the X-Sub benchmark with 4.46 M fewer parameters than GAT, one of the best SOTA methods. Inference speed of our model reaches as fast as 86.23 sequences/(second × GPU). Extensive experimental results further validate the effectiveness of our proposed approach on three large-scale datasets, namely, NTU RGB+D 60, NTU RGB+D 120, and Kinetics Skeleton.

Due to the outbreak of lung infections caused by the coronavirus disease (COVID-19), humans have to face an unprecedented and devastating global health crisis. Since chest computed tomography (CT) images of COVID-19 patients contain abundant pathological features closely related to this disease, rapid detection and diagnosis based on CT images is of great significance for the treatment of patients and blocking the spread of the disease. In particular, the segmentation of the COVID-19 CT lung-infected area can quantify and evaluate the severity of the disease. However, due to the blurred boundaries and low contrast between the infected and the non-infected areas in COVID-19 CT images, the manual segmentation of the COVID-19 lesion is laborious and places high demands on the operator. Quick and accurate segmentation of COVID-19 lesions from CT images based on deep learning has drawn increasing attention. To effectively improve the segmentation effect of COVID-19 lung infection, a modified UNet network that combines the squeeze-and-attention (SA) and dense atrous spatial pyramid pooling (Dense ASPP) modules) (SD-UNet) is proposed, fusing global context and multi-scale information. Specifically, the SA module is introduced to strengthen the attention of pixel grouping and fully exploit the global context information, allowing the network to better mine the differences and connections between pixels. The Dense ASPP module is utilized to capture multi-scale information of COVID-19 lesions. Moreover, to eliminate the interference of background noise outside the lungs and highlight the texture features of the lung lesion area, we extract in advance the lung area from the CT images in the pre-processing stage. Finally, we evaluate our method using the binary-class and multi-class COVID-19 lung infection segmentation datasets. The experimental results show that the metrics of Sensitivity, Dice Similarity Coefficient, Accuracy, Specificity, and Jaccard Similarity are 0.8988 (0.6169), 0.8696 (0.5936), 0.9906 (0.9821), 0.9932 (0.9907), and 0.7702 (0.4788), respectively, for the binary-class (multi-class) segmentation task in the proposed SD-UNet. The result of the COVID-19 lung infection area segmented by SD-UNet is closer to the ground truth compared to several existing models such as CE-Net, DeepLab v3+, UNet++, and other models, which further proves that a more accurate segmentation effect can be achieved by our method. It has the potential to assist doctors in making more accurate and rapid diagnosis and quantitative assessment of COVID-19.

In the medical field, Alzheimer’s disease (AD), as a neurodegenerative brain disease which is very difficult to diagnose, can cause cognitive impairment and memory decline. Many existing works include a variety of clinical neurological and psychological examinations, especially computer-aided diagnosis (CAD) methods based on electroencephalographic (EEG) recording or MRI images by using machine learning (ML) combined with different preprocessing steps such as hippocampus shape analysis, fusion of embedded features, and so on, where EEG dataset used for AD diagnosis is usually is large and complex, requiring extraction of a series of features like entropy features, spectral feature, etc., and it has seldom been applied in the AD detection based on deep learning (DL), while MRI images were suitable for both ML and DL. In terms of the structural MRI brain images, few differences could be found in brain atrophy among the three situations: AD, mild cognitive impairment (MCI), and Normal Control (NC). On the other hand, DL methods have been used to diagnose AD incorporating MRI images in recent years, but there have not yet been many selective models with very deep layers. In this article, the Gray Matter (GM) Magnetic Resonance Imaging (MRI) is automatically extracted, which could better distinguish among the three types of situations like AD, MCI, and NC, compared with Cerebro Spinal Fluid (CSF) and White Matter (WM). Firstly, FMRIB Software Library (FSL) software is utilized for batch processing to remove the skull, cerebellum and register the heterogeneous images, and the SPM + cat12 tool kits in MATLAB is used to segment MRI images for obtaining the standard GM MRI images. Next, the GM MRI images are trained by some new neural networks. The characteristics of the training process are as follows: (1) The Tresnet, as the network that achieves the best classification effect among several new networks in the experiment, is selected as the basic network. (2) A multi-receptive-field mechanism is integrated into the network, which is inspired by neurons that can dynamically adjust the receptive fields according to different stimuli. (3) The whole network is realized by adding multiple channels to the convolutional layer, and the size of the convolution kernel of each channel can be dynamically adjusted. (4) Transfer learning method is used to train the model for speeding up the learning and optimizing the learning efficiency. Finally, we achieve the accuracies of 86.9% for AD vs. NC, 63.2% for AD vs. MCI vs. NC respectively, which outperform the previous approaches. The results demonstrate the effectiveness of our approach.